zoom-video.rb

def self.generate_video(file,idx)
    ipl_image = IplImage::load(file.path)
    rm_image = Image.read(file.path).first

    op = CvHaarClassifierCascade::load('haarcascade_frontalface_alt.xml')

    chosen = op.detect_objects(ipl_image).to_a.sample

    raise "no face found" if chosen.nil?

    face_center_x = chosen.center.x
    face_center_y = chosen.center.y


    height = rm_image.rows
    width = rm_image.columns
    og_height = height
    og_width = width

    avg_scale_factor = 50 * ((chosen.width.to_f / width) + (chosen.height.to_f / height) / 2)
    step = [avg_scale_factor / 300, 0.0005].max

    if height > width
      height = [600, height + (height % 2)].min
      width = (width.to_f * height.to_f / og_height).to_i
      width = width + (width % 2)
    else
      width = [800, width + (width % 2)].min
      height = (height.to_f * width.to_f / og_width).to_i
      height = height + (height % 2)
    end
    scale = "#{width}:#{height}"

    out_video_path = "output/output-#{idx}.mp4"
    audio_library_path = 'sounds/song1'
    audio_path = Dir.glob("#{audio_library_path}/*.mp3").sample
    input = "-i #{file.path}"
    audio_input = "-i #{audio_path}"
    vid_params = "-c:v libx264 -r 30 -pix_fmt yuvj420p"
    command = "/Users/casey/bin/ffmpeg -y -loglevel fatal -loop 1 #{input} #{audio_input} #{vid_params}"
    zoom_speed = step
    x_comp = "'iw*#{face_center_x/og_width}-(iw/zoom/2)'"
    y_comp = "'ih*#{face_center_y/og_height}-(ih/zoom/2)'"
    pan = "x=#{x_comp}:y=#{y_comp}"
    zoompan = "-vf \"scale=iw*2:ih*2,zoompan=z='zoom+#{zoom_speed}':d=500:#{pan},scale=#{scale}\""
    duration = "-t 10"
    input_map = "-map 0:0 -map 1:0"
    `#{command} #{zoompan} #{duration} #{input_map} #{out_video_path}`
    out_video_path
  end